Media fingerprinting provides an ability to link short, unlabeled, snippets of media content to corresponding data. It provides the ability to automatically identify and cross-link media, such as audio and video. Unlike many competing technologies, a goal of media fingerprinting is to perform recognition without imposing extraneous hardware restraints to automatic detection and/or replacement, as well as without extraneous data transmission Fingerprints or descriptors can then be derived from features computed over the reference samples. Fingerprints of a media item can then be compared to fingerprints of reference samples to determine identity of the media item.
Various challenges are posed when systems do not function with exact bit-level matches when comparing content. The system may be functioning with multiple media items that are identical or nearly identical. Because media items can comprise varied content but relate to a common media item, multiple fingerprints are stored. Additionally, there are difficulties introduced through the numerous forms of playback available to the end consumer. Media that is played through a cell phone, computer speakers, or high-end media equipment will have very different characteristics.